Stephen Baker

The Numerati
Home - posts tagged as Boost

Google+ and the Numerati
July 11, 2011Boost

The first circle I filled on Google+ was easy. Family. I dragged and dropped different family members into the circle, and I guess some day I could send them all an update or a photo. Family's easy, because it's defined. But... Are my in-laws interested in the same stuff as my sisters and my kids? And won't nieces and nephews be bored by some of the middle-aged traffic? Hmmm. Maybe I should have subgroups in family.

Then comes work. What do my old colleagues from the papers in Caracas and El Paso have in common with my BusinessWeek friends? They're journalists. That's something. But if one of my messages is BW-specific, they probably won't be interested. Then there are divisions within BW itself. I could think of it this way: 1) the people I had lunch with, 2) the people I probably should have had lunch with, and 3) those I'd never have lunch with. Circles and more circles.

Grouping people is difficult. This is what the Numerati deal with all the time. Should Mercedes buyers be grouped with country-club members or European vacationers? The quants work out the numbers. But there's a key difference. If a Mercedes buyer who hates golf gets a come-on from a country club, he's not offended. It's just one more banner on the screen or piece of junk mail to throw out.

Friendships don't work that way. If I send out a mailing that is appropriate for 78% of the people in a certain circle, 22% of my contacts feel that I'm treating them as names on a bulk mail list. And they're right.

So what's the answer? Google would love it if each one of us would carefully delineate our friendships in hundreds of small and overlapping circles. They want us to lay out our social grafs in exquisite detail. That way they can go to school on our relationships and teach their machines to target our friends with the same precision that we're striving for.

I don't think I'll bother. The more I analyze my friends, the more differences pop up among them. This isn't to say that I won't send them all a promotion for my next book. But that's a bulk mailing for publicity and labeled as such. However, it's really hard--impossible, I'd say--to fit friendships into boxes (or circles). I can imagine forwarding one article that would appeal to three former BusinessWeek colleages, two from Caracas, my Spanish-speaking nephew and two friends from Spain. I'll address it manually. I see no other way. 

link to post share:

Definition of an AOL searcher
February 6, 2010Boost

The users of different search engines fall into different tribes, according to this study in Ad Age. (ex Unbound Edition) That shouldn't come as a surprise. I found the definition of the AOL searcher especially poignant:

AOL customers feel less intellectual than their peers, are 55 and older, spend their money more responsibly, want to blend in to the crowd, feel like they've gotten a raw deal out of life, expect less from their future and, believe it or not, still use dial-up modems.

Hmmm. What sort of marketing campaign should we work up for them?

Incidentally, if any of you speak German, here's an article I wrote for the Frankfurter Allgemeine Zeitung. It's about what we should be storing in our heads, a theme I've been exploring a bit in recent talks.

add comment link to post share:

X-box Live: Where does it fit into "friends" series?
April 25, 2009Boost

Talked the other day with Marc Whitten, general manager of Microsoft's Xbox division. This was for my BusinessWeek reporting on Friendship. Xbox seemed relevant, because 20 million people play on the live version, most of them searching, meeting, and playing with friends of one type or another.

There are two approaches to friend-finding on Xbox. One is to find people all over the world to compete with. That was the initial focus, Whitten says. But his team has learned that a larger and richer market features people who just want to hang out with friends they already have. They may be down the block or across an ocean from each other.

In these meetings,  the players often have unequal skills. This has led to a growth in non-competitve games. People play on the same team. Whitten does this with his brother in Oklahoma. Otherwise, he says, he'd blow him away. After all, the guy lives Xbox.

I was hoping to learn that Microsoft was crunching all the Xbox Live conversation and friending data to find out about hierarchies and networks of friends. That's not the focus, though. They're less interested in learning about people than simply helping them find and play with their friends.

add comment link to post share:

danah boyd on Friendship
April 12, 2009Boost

This is a little film clip I did with danah boyd, a Microsoft researcher. She talks about decoding the patterns of our friendships. (If I were a pro, I would have edited out the first couple of seconds before she realizes the camera's running. But, actually, that moment of recognition, when she snaps into a smile, is my favorite moment in the whole clip. Not that you shouldn't watch the rest...)

This is part of the Friendship series I'm working on at BusinessWeek. I'm heading out to the West Coast tomorrow, and hope to talk to people at Facebook, LinkedIn, and elsewhere. I'll take along my flip camera and post more video, too.

add comment link to post share:

Locaccino: Tracking location with privacy
March 2, 2009Boost

Locaccino is a new location-based friend-finding service for Facebook. It comes out of Carnegie Mellon University, and the emphasis is on customizable privacy settings.

From the Web page:

Our technology allows you to easily define the times when you want to share your locations. Locaccino lets you create groups for your friends to simplify your location preferences. With Locaccino, allowing your co-workers to see your location from 8AM – 5PM is a snap.

Locaccino also allows you to specify where you can be located. Using a Google maps interface, you can define regions where you do and don’t want other people to be able to find you.

We'll see if Google adds these features to its Latitude application (which I have on my Blackberry). Of course, part of the fun is seeing what my colleagues are up to over the weekend. But we all should have the right to draw the curtain when we want. (If my editor looks on his Blackberry today, he'll see my dot blinking in snowy Montclair. Why suffer two monster commutes to New York just to write? At least that's my excuse.)

The view from my front window

add comment link to post share:

Microsoft search puts us in groups
January 30, 2009Boost

Search has hit a plateau, according to Microsoft researchers. Figuring out the most relevant Web page based solely on the search query isn't advancing much beyond the state-of-the-art established in the last decade by Google. (And common wisdom is that Microsoft and Yahoo have caught up in quality, if not speed.)

Now Microsoft researchers, according to this story in MIT Technology Review, are attempting to group Internet searchers with others who have similar patterns. The idea is that we search for the same things as members of our behavioral tribe, and that with this added analysis, the search engines will deliver more relevant results. (They're going to present their study at a search gathering in Barcelona next month.)

One challenge will be gathering the data. For their study, researchers used a group of 100 colleagues who had agreed to participate. More details:

The researchers grouped people using explicit factors, such as their age, gender, participation in certain mailing lists, and job function. In some cases, implicit groups--such as people who appeared to be conducting the same task or appeared to have the same interest--were inferred. The researchers acknowledged that gathering such data in the real world could be tricky. But it could perhaps be collected through registration, by caching previous searches or by tapping into social-networking software.

add comment link to post share:

How the Numerati helped win it for Obama
January 24, 2009Boost

An essay I wrote:

News Analysis January 23, 2009, 6:35PM EST

What Data Crunchers Did for Obama

Sophisticated political microtargeting efforts are grouping us in surprising ways. For Obama, swing voters known as Barn Raisers proved pivotal

About three minutes into his speech on Jan. 20, President Barack Obama spoke a word never before uttered in a Presidential inauguration speech: "data.". The word may sound nerdy, and Obama used it in reference to indicators of economic and other crises. But it's no coincidence the word found its way into his remarks. The harnessing of data has been crucial to Obama's rise to power.

Throughout the campaign, Obama and his team not only bested his Democratic and Republican rivals in social networking and fund-raising through the Internet, they also engaged in a data battle to locate potential swing voters. These efforts zeroed in on hotly contested states and congressional districts, where the shift of 1,000 or 2,000 voters could prove decisive—meaning the focus was on only a tiny fraction of the voting public. But to find those swing voters, both sides hired tech wizards to sift through mountains of consumer and demographic details. They scrutinized nearly everyone they could find.

Ten "Tribes"

One Democratic consultancy, Spotlight Analysis, took this hunt to extraordinary lengths. Working on behalf of Democratic candidates, though not directly for the Obama campaign, Spotlight crunched neighborhood details, family sizes, and purchasing behavior. It then grouped nearly every American of voting age—175 million of us—into 10 "values" tribes. Fellow tribe members may not share the same race or religion, or fall into the same income bracket, but they have common feelings about issues that transcend politics: God, community, responsibility, and opportunity. Spotlight believes that one of these tribes, a morally guided (but not necessarily religious) grouping of some 14 million voters—dubbed "Barn Raisers"—held the key to the contest between Obama and his Republican challenger, Arizona Senator John McCain.

The definition of a Barn Raiser cuts straight to the heart of what distinguishes political microtargeting from traditional political groupings. Barn Raisers can be of any race, religion, or ethnic group. About 40% of Barn Raisers are Democrats, or lean that way, and 27% favor Republicans—though the group strongly supported President Bush in his 2004 reelection campaign. Barn Raisers are slightly less likely to have a college education than Spotlight's other swing groups. They're active in community organizations but are ambivalent about government. And they care more deeply than most people about "playing by the rules" and "keeping promises," to use Spotlight's definitions.

With special appeals to Barn Raisers in swing states, Spotlight's clients, including the Service Employees International Union and the Democratic Senatorial Campaign Committee, hoped to turn battleground states such as Florida, North Carolina, Virginia, and Ohio. The data-based techniques they put to use, similar to those used to target supermarket shoppers and even to hunt for terrorists, are turning politics into the sophisticated calculations typically associated with Google (GOOG) and its ilk. In a fraction of a second, computers sort us into segments and then calculate the potential that each of us has to swing from red or purple to blue. For many, this signals the dehumanization of politics.

Others say political data mining helps better pinpoint individuals whose views and priorities may otherwise be overlooked. Consider a voter in, say, Richmond, Va. Republican and Democratic data miners count the number of children she has in school, they take note of her car, her Zip Code, her magazine subscriptions, and the balance on her mortgage. They might even find in her data that she has two cats and no dog. (Cat owners lean slightly for Democrats, dog owners trend Republican.) In the end, they place her into a political tribe and draw conclusions about the issues that matter to her. Is that so horrible?

Behavioral Grouping

For generations, politicians lacked the means to study us as individuals. So they placed us into enormous groups—blacks, Jews, gays, union members, hunters, soccer moms—and treated us as masses. While the rich and well-connected got to collar candidates at $1,000-a-plate dinners, the rest of us were processed as herds.

Nowadays, Spotlight and other microtargeters (for both parties) continue to place us into big groups. But the divisions are based more on our behavior and choices, and less on the names, colors, and clans that marked us from birth.

Spotlight embarked on its research three years ago by interviewing thousands of voters the old-fashioned way. At first, Barn Raisers didn't seem especially noteworthy. The group represented about 9% of the electorate. It spanned genders, races, and religions.

But when Spotlight's analysts dug deeper, they discovered that Barn Raisers stood at the epicenter of America's political swing. In 2004, 90% of them voted for President Bush, but then the group's political leanings shifted, with 64% of them saying they voted for Democrats in the 2006 election. Spotlight surveys showed that political scandals, tax-funded boondoggles like Alaska's Bridge to Nowhere, and the botched job on Hurricane Katrina sent them packing.

Suddenly, Spotlight had a line on millions of swing voters. The challenge then was to locate groups of them in swing states. For this, the company analyzed the demographics and buying patterns of the Barn Raisers they surveyed personally. Then it instructed its computers to scour commercially available databases for others with matching profiles. By Spotlight's count, this approach nailed Barn Raisers three times out of four. So Democrats could bet that at least three-quarters of them would be likely to welcome an appeal stressing honesty and fair play.

Still Swing Voters

Did it work? Spotlight hasn't yet carried out the surveys to determine how many of its Barn Raisers backed Obama. But it's reasonable to presume that amid that sea of humanity stretched out before Obama on Washington's Mall on Jan. 20, at least some of those were moved by microtargeted appeals. And if Obama and his team fail to honor their mathematically honed vows, the Barn Raisers may abandon them in droves. They're swing voters, after all.

And if there is one thing the research has made clear, it's this: Even if Barn Raisers exist as a tribe only in a database, they take broken promises very seriously. And they probably won't object if data-mining politicians figure that out.

add comment link to post share:

How friends' friends can affect moods
January 3, 2009Boost

One of my Twitter buddies sent me the link to this New Scientist story on how things like happiness work their way through social networks. It's not just your friend's happiness that influences you, but also your friends' friends. So if you're unhappy, or thinking about taking up smoking, or even concerned that your child might be diagnosed with autism, look around and see what your friends' friends are up to.

Researchers are going to benefit from lots of new data about our social connections in the next decade or so. Facebook and MySpace are gold mines, of course. But there are also companies such as Sense Networks that are starting to track our physical movements, and to sort us into new behavioral tribes.

Not really a social network or a web. It's just a leaf I found last summer. But you get the idea...

Soon, researchers like those mentioned in the article will be able to study the contagion patterns of obesity, happiness, suicide, political philosophy. And of course the government will be especially keen to understand the migration across social networks of terrorist sympathies.

When I was researching The Numerati, I talked to scientists at the Naval Postgraduate School in Monterey, Calif., who are attempting to model the diffusion of certain ideas in places like Iraq. (Or, more specifically, in Iraq.) Let's say the United States does something good in a town, like building a medical clinic. How does positive news spread to other towns? When I was there, the studies were based on simulations using agent-based modeling. In the future, they'll be able to replace many of these simulations with reality: our behavior.

add comment link to post share:

Modeling the cheating husband
November 26, 2008Boost

I swear I was working (on this Thanksgiving eve) when I came across this Cosmopolitan article on how to spot the cheating husband. It caught my attention because the instruction form a primitive version of a predictive mathematical model. In a sense, this model of a (potentially) cheating husband resembles the models Fair Isaac builds of people likely to default on mortgages.

The Private Life of San Juan (1934)

And, suffice to say, both models, Fair Isaac's and Cosmos, can be wrong.

While Fair Isaac studies how many credit cards we have and how we juggle financial obligations, Somos focuses on different variables:

  • Was he spoiled as a kid?
  • Do his parents tend to baby him and help him out of financial jams?
  • Has he ever bragged about cheating on an exam or paying someone to write a paper for him in college?
  • Does he work mostly with women?
  • Is he always logging in late hours, whether it be at the office, at dinner with clients or on business trips?
  • Does he make a lot of money?
  • Can he talk his way out of anything (parking tickets, rolling into work late)?
  • Does he make an effort to charm everyone -- your coworkers, your older sister, a saleswoman?
  • When you go to parties, does he insist on making the rounds?
  • Does he usually hang out with a crew of mostly single guys?
  • Do his friends encourage him to join them in just-for-men activities?
  • Do his pals have problems staying in relationships?

  • If the Numerati were to delve into this type of data, of course, they'd have to come up with a weighting for each variable. They'd build the model from a population including known cheaters, and then test it on others. Somehow I doubt Cosmo goes to such lengths.

    add comment link to post share:

    Found my grocery store bucket (or one of them)
    October 21, 2008Boost

    This data mining post  declares that the oft-cited correlation between beer- and diapers-buyers is an urban legend. But they do unearth grocery data indicating that there one bucket of people who buy cat food, chocolate and red wine.

    Hey, I thought: That's us! But the researchers clearly didn't get their data from New Jersey, because we buy our chocolate and cat food in one store and then have to go to another to buy wine (of any color).

    add comment link to post share:

    ©2021 Stephen Baker Media, All rights reserved.     Site by Infinet Design

    Kirkus Reviews -

    LibraryJournal - Library Journal

    Booklist Reviews - David Pitt

    Locus - Paul di Filippo

    read more reviews

    Prequel to The Boost: Dark Site
    - December 3, 2014

    The Boost: an excerpt
    - April 15, 2014

    My horrible Superbowl weekend, in perspective
    - February 3, 2014

    My coming novel: Boosting human cognition
    - May 30, 2013

    Why Nate Silver is never wrong
    - November 8, 2012

    The psychology behind bankers' hatred for Obama
    - September 10, 2012

    "Corporations are People": an op-ed
    - August 16, 2011

    Wall Street Journal excerpt: Final Jeopardy
    - February 4, 2011

    Why IBM's Watson is Smarter than Google
    - January 9, 2011

    Rethinking books
    - October 3, 2010

    The coming privacy boom
    - August 17, 2010

    The appeal of virtual
    - May 18, 2010